The guild was busier in the afternoon than it had been in the morning.
More people at the tables, a longer queue at the counter, the noise level higher in a way that felt less like activity and more like the particular restlessness of people waiting on things they could not control. He joined the queue and waited, turning the quest posting over in his hand, and watched the room the way he had been watching everything since he arrived in this world, cataloguing without engaging.
His compounding panel had been open in the corner of his vision for the entire walk back from the cave. He had not been able to stop checking it. Combustion was sitting at 1,389.2 VP now and the rate at which that number moved had continued to accelerate in a way that he could actually perceive in real time if he watched it for more than thirty seconds. It was not the number itself that he kept looking at. It was the speed of it. The old rate had been something he checked periodically and found marginally higher. The corrected rate was something he could watch change.
He reached the counter. The same woman from the morning looked at him, looked at the quest posting he placed on the counter and then looked at his token. She checked something on her side, made a notation and produced a small envelope that she set beside the posting. He picked it up and opened it. Coins, a mix of denominations he could not read the values of yet but could estimate by size and weight relative to each other. He counted them into his palm and closed his hand around them.
His first earned currency in Erasval. He had no frame of reference for what it was worth but the weight of it felt sufficient for a twenty-kill quest at his rank. He pocketed it and stepped away from the counter.
He found the same corner table he had used that morning and sat down.
He was not going to spend the money yet. He did not know prices, did not know what a fair exchange looked like, did not know which merchants in this settlement were trustworthy and which were not. Spending currency before he understood its value was the kind of mistake that only happened once if you were paying attention, and he was paying attention.
He pulled up the Compound System instead and looked at the investment panel properly for the first time since the error correction.
The numbers were still moving.
Ignition: invested 20 VP, return value 44.6 VP. Time invested: 20 hours.
Heat Control: invested 700 VP, return value 891.4 VP.
Combustion: invested 1,000 VP, return value 1,391.7 VP.
Spatial Awareness: invested 300 VP, return value 362.8 VP.
Ashveil Blade: invested 500 VP, return value 509.3 VP.
He looked at Ignition. Twenty VP in, return value sitting at 44.6, which meant if he withdrew right now the VP was gone but the skill itself would be upgraded by whatever the system calculated from twenty hours of compounding on a 20 VP base. The return value was the measure of how much growth had accumulated, not a refund. Withdrawing did not give it back. It spent it, permanently, on the upgrade.
He had been thinking about this since the error correction had clarified the mechanic to him. The tension in the system was not complicated once you saw it clearly. Longer investment meant more compounding meant better upgrade on withdrawal. But withdrawing reset the clock entirely. The asset went back to zero investment, zero compounding, and had to be restarted from scratch. Whatever he had built up was crystallised into the upgraded version of the skill or item and the accumulation was over.
Which meant the question was never whether to withdraw. It was always when.
He looked at Ignition again. Twenty VP over twenty hours had produced a return value of 44.6, meaning the growth was 24.6 VP of compounded value sitting in the skill waiting to be realised. Not a large number. But he was not withdrawing Ignition because he expected a dramatic result. He was withdrawing it because he needed to understand exactly what withdrawal produced before he made any decisions about the larger investments.
He selected Ignition and opened the withdrawal prompt.
[Withdraw investment from: Ignition (F) - Proficiency 1]
[Current return value: 44.6 VP]
[Investment VP: consumed on withdrawal. No VP returned.]
[Asset will be upgraded based on accumulated return value.]
[Upgrade threshold analysis:]
[Return value 44.6 VP on F-grade asset — Grade increase threshold: not met (requires 100 VP return value).]
[Power increase: proportional to return value accumulated. Estimated output increase: 34%.] [Confirm withdrawal?]
He read the threshold line twice. Grade increases required 100 VP return value on an F-grade asset. He had 44.6. Not enough for a grade jump, which meant what he was getting was a power increase only, no letter grade change, no proficiency movement. Just Ignition hitting harder by approximately a third.
He confirmed it.
The skill entry updated immediately.
[Ignition (F) - Proficiency 1]
[Status: Standard — no active investment]
[Output: +34% from base]
He looked at it for a moment. F-grade still. Proficiency 1 still. But the output modifier sitting underneath it was new, a permanent change baked into the skill that would persist regardless of what he did next. The base had shifted upward. Whatever Ignition produced now it produced 34% more of it, and that modifier would stack with any future investment withdrawal on top of it.
He thought about what 44.6 VP of return value had bought on a skill he had barely invested in. Then he thought about what 1,391.7 VP of return value would buy on Combustion if he left it running for another week before withdrawing.
He did not touch Combustion.
He opened the Ignition investment prompt again immediately.
[Invest in: Ignition (F)]
[Amount to invest:]
The debate took longer than he expected. The whole point of the withdrawal had been to test the mechanic, and now that he understood it the question of what to reinvest was not trivial. The grade threshold was 100 VP return value. To reach 100 VP return value he needed enough base investment running long enough for the corrected exponential rate to push the return value past that number. On 20 VP it had taken twenty hours to reach 44.6. On a larger base the rate would accelerate faster. The exponential scaling rewarded heavier investment and longer hold times simultaneously.
He typed 150.
[Confirm investment of 150 VP into Ignition (F)?]
Confirmed.
The bar appeared. Wider than before, the larger base giving it more to work with from the start. He watched it pulse twice and then pulled his attention away from it deliberately. Watching it was not useful. Leaving it alone was.
He leaned back in the chair and looked around the guild floor, the coin purse sitting in his pocket, the Ashveil Blade at his side, four investments running in the background at a rate that was going to look very different in twenty four hours than it did right now.
The language assimilation counter sat at 11% in the corner of his vision.
He had been in Erasval for approximately thirty six hours.
He ordered something from the person moving between the tables carrying food, using the only real currency that worked without language, pointing at what the person two tables over was eating and holding up one finger. The food arrived. He did not know what it was. It was hot and substantial and he ate all of it.
Then he sat with his hands around a cup of something warm that was not quite anything he had a name for and thought about the grade threshold he had just read.
One hundred VP return value for an F-to-E grade jump. He did not know the threshold for E-to-D, or D-to-C, or what happened when a skill hit S-grade and kept going. He did not know whether higher grade assets had higher thresholds or whether the thresholds scaled with the grade of the skill at the time of withdrawal. He did not know whether there was a ceiling on how many times the same asset could be upgraded before the system considered it fully realised.
He had one data point. F-grade, 44.6 VP return, 34% power increase, no grade change.
He needed more data points and the only way to get them was to run more investments for longer and withdraw at different return value thresholds and record what happened.
He pulled up a blank notation function in the system panel he had found earlier in the day, a simple text field that appeared to exist for exactly this kind of record keeping, and wrote down what he knew.
[Ignition withdrawal, return value 44.6, result: +34% power, no grade change, no proficiency change. Grade threshold F to E, minimum 100 VP return value required.]
Two lines. The beginning of something that was going to get much longer.
Combustion's return value ticked to 1,412.3 VP in the corner of his vision.
He closed the notation panel and finished his drink.